Despite strong AI translation capabilities, the platform’s workflow made reviewing and editing translated content slow and fragmented. I redesigned key interaction patterns to help teams validate and refine AI output more efficiently.
Role
Product Designer
Team
Product Designer (Me)
2 Designers
UX Researcher
Product Manager
2 Developers
Timeline
Aug–Dec 2024
Tools
The Problem
Vosyn builds AI-powered tools that help global teams translate and localize content across multiple languages. During my internship, I worked on improving workflows across several internal products used for AI-assisted translation and review.
The system could generate translations quickly, but reviewing and refining those outputs required navigating several disconnected interfaces. My work focused on redesigning those workflows so teams could validate and edit AI-generated content more efficiently.
Because of NDA restrictions, I can only share redacted product interfaces or detailed design artifacts from this project.
Outcomes
64 → 82
increase in System Usability Score after usability testing
28%
reduction in multi-step task time during localization workflows
5
AI-assisted workflow features prototyped and tested to improve translation review
These improvements helped teams validate machine-generated translations faster and reduced friction across the localization process.
The Journey
The platform already had strong AI translation capabilities, but usability testing revealed that most friction happened after generation. Users struggled to compare original and translated content and often had to move between multiple panels just to review a single translation.
I conducted usability sessions and workflow analysis to understand where users lost time and context. A consistent pattern emerged: translation generation was fast, but validation was slow.
Based on these findings, I explored interaction patterns that kept generation, review, and editing within a single working context. Prototypes focused on reducing navigation overhead and helping reviewers identify where AI output required attention.
Reflection
This project reinforced an important lesson about AI products: generating content is only part of the workflow. Users also need efficient ways to evaluate, correct, and trust machine-generated output.
Small changes to workflow structure had a larger impact than adding new features. By keeping users in a single review context and reducing the cost of verification, the platform became significantly easier to use for teams working with multilingual content.